Open dmadisetti opened 4 years ago
We will use the 3D model and csv data to generate the mask of the vehicle of interest on each image. We should all be looking at this first before we can make any progress since it builds the foundation for our pipeline.
Some more reading https://arxiv.org/pdf/1806.05034.pdf
Check out prior art (the notebooks give good examples):
Predict Distances
There are also unsupervised segmentation techniques